How do we identify our fellow human beings? This is a seemingly simple question, but one that confounds scientists across an array of disciplines, especially in computer animation and robotics. This problem was distilled down to the idea of the “uncanny valley” by roboticist Masahiro Mori.
This image describes the uncanny valley. On the x-axis we have a hypothetical measure of human likeness, that is, how much something physically resembles a human. On the y-axis we have a measure of how attached or familiar we are with the object. As human likeness increases, so too does our attraction to the object.
Compare your desk lamp with the animated lamp from the Pixar, named Luxo Jr. The animated lamp has taken on several human characteristics, evidenced through its body and head motion. Moving farther along the graph of human likeness we arrive at something along the lines of the Brave Little Toaster or Johnny 5 both of which are given human facial expressions. Still further we have human-like robots such as C3P0 who is human in form and manner, but is still clearly a robot. A step further and our graph inverts on itself.
Did You Know?
The first robot was a die-casting robot built for General Motors. It is now on display at the Smithsonian in Washington.
As we get closer to human likeness, there is a stage were we stop seeing inanimate objects which have human characteristics and we start seeing things that are human in appearance but alien in behavior. Our familiarity with these characters plummets to the point where we find them repulsive or disturbing.
Two famous examples exist in computer animation, The Polar Express and the baby in Pixar’s Tin Toy. Tin Toy is probably the best example of this principle. The Tin Toy itself has human features, but is clearly not human. His expressions and animation endear him to us. The baby on the other hand, clearly is human but its movement is somehow wrong, making it very unsettling.
Did You Know?
TRON was a trailblazer in bringing computer-generated imagery to movies. The computers used to render the graphics in TRON had 2 MB of memory and 330 MB of storage.
As we further increase human likeness, we start having characters that are, for all intents and purposes, indistinguishable from real humans. We start to relate to them as we would a fellow human. One of the more advanced computer animated humans is Emily.
The uncanny valley is fairly well accepted though the theoretical basis for it is still very much unknown. We may feel unsettled by near-human characters because they remind us of corpses or because their unnatural behavior triggers a deep-seated survival instinct which forces us to distance ourselves from them.
Many clever companies skirt around the issue by making endearing characters, which fall on the near side of the valley. Pixar, for example, usually creates human characters with cartoonish, exaggerated features. They are extremely expressive and lovable, and they don’t trigger the sense of unease that accompanies the uncanny valley.
Robots have come a long way, but even the best human-like robots still tend to fall squarely into the uncanny valley. We're some years away from the very realistic robots of AI, and even further from the indistinguishable Cylons, but robotics is still a massive focal point for research. We are starting to see some interesting and sophisticated robots doing everything from elderly care to biotechnology research. Understanding the uncanny valley will have a huge impact on how we interact with the robots of tomorrow.
Slate Magazine: The Undead Zone – Why realistic graphics make humans look creepy
Popular Mechanics: The Truth About Robots and The Uncanny Valley
Article first published May 5, 2011